software engineers

Abstract: Diffusion Language Models (DLMs) offer promising parallel generation capabilities but lag behind autoregressive models in complex reasoning and tool-use tasks. While Reinforcement Learning (RL) has recently been applied to enhance DLMs, standard RL approaches suffer from an exploration bottleneck. To address this, we inject reasoning priors from a stronger teacher model to guide RL exploration. In this paper, we introduce CanvasAnneal, a curriculum-guided diffusion RL framework. During the initial RL phase, we warm-start exploration by injecting teacher-generated reasoning traces into the initial diffusion canvas. As training progresses, we gradually remove this guidance and require the model to generate more of the reasoning trajectory independently. Across mathematical reasoning and tool-use benchmarks, CanvasAnneal improves over standard diffu-GRPO on MATH500, Countdown, and Tau2 and substantially accelerates reward improvement on several tasks, while gains are task-dependent. Our results suggest that structured training-time guidance can alleviate exploration bottlenecks in diffusion RL and speed up convergence on harder tasks.

Abstract: Large language model agents are expected to continuously adapt to new tasks and environments over their lifetime by reusing past experience. However, existing memory-based agents struggle to transfer reusable experience across environments and suffer from catastrophic forgetting as experience accumulated. To address these challenges, we propose LifeMem, a lifelong learning framework that enables agents to transfer knowledge across multiple environments. During learning, LifeMem clusters accumulated interaction trajectories based on underlying workflows to extract reusable skills. When solving a new task at inference time, the agent recalls relevant skills and trajectories to guide actions. To validate our method, we conduct experiments across 10 environments and over 13k tasks with 2k newly annotated interaction trajectories. Results show that LifeMem enables effective experience reuse in lifelong learning, achieving both reduced forgetting on learned tasks and superior cross-task transfer. Further analysis reveals that task streaming impacts learning, while consolidating structurally similar trajectories within memory boosts performance.

Abstract: Even with a fixed dataset and research question, data analysis involves many defensible decisions. Understanding how these choices influence the results is scientifically important but remains challenging. Crowdsourcing and agentic AI can generate hundreds of end-to-end analyses, but scaling generation alone can create a processing bottleneck and an analytic ``black hole.'' A common workaround is to impose a shared fixed decision taxonomy, which can limit insight and understate uncertainty. We present ForkSCOPE, a human-AI collaboration framework that induces structure bottom-up from the code corpus of end-to-end analyses, without a taxonomy fixed before or after generation, so the organization and evaluation of the garden can scale with the corpus. ForkSCOPE surfaces the charted garden of forking paths through a human-AI collaboration pipeline and an evidence-linked interactive viewer for steering and verification: it spotlights organically identified forks and structures and produces a derived taxonomy and decision map compatible with existing multiverse tools.

Abstract: We prove a lower bound of Ω(log n/ log log n) on the approximation ratio of the global grammar-based compression algorithm Greedy. To our knowledge, the previously best lower bound was a constant, and the existence of a nonconstant lower bound had remained open for more than twenty years. Our bound holds on an infinite family of words of length n, over alphabets of growing size, for every execution using left-to-right occurrence replacement and arbitrary tie-breaking. The lower bound is also formally verified in Lean 4.

Abstract: The sparse matrix-vector multiplication (SpMV) algorithm is a fundamental computational kernel of linear algebra and serves as a building block for numerous applications, primarily iterative solvers for systems of linear equations used in scientific and engineering simulations. This paper compares vectorized implementations of the SpMV algorithm across eight established sparse matrix storage formats and proposes a novel modification of the CSR format, Hierarchical CSR (HCSR), which enhances SpMV performance on RISC-V processors. Our SpMV implementations utilize RVV 1.0 intrinsics and are publicly available as an open-source C++ library named RVVLASparse. Computational experiments conducted on SpacemiT K1 and K3 RISC-V boards demonstrate that selecting an appropriate matrix storage format accelerates SpMV computations by an average of 1.6x, while the proposed HCSR format achieves the shortest execution time among all considered formats across a broad class of sparse matrices.

Abstract: We present our system for the SHROOM-Visions 2026 shared task on character-level VLM hallucination detection. A small ($4$B-parameter) VLM is fine-tuned as a per-token classifier reading a two-token feature from its own hidden states, and is ensembled with a $\sim$400B zero-shot VLM judge at prediction time. Both components see off-the-shelf OCR of any visible in-image text. We use synthetic hallucination data generated by the large model as a source of ensemble diversity, and use validation to select feature layer, training data and OCR grounding. Our official entry reaches mean Cor $0.487$ / Cor-lbl $0.387$ on the hidden test set, placing $6$th/$28$ (EN), $6$th/$21$ (FR), $8$th/$21$ (IT) and $7$th/$22$ (ZH) on the task's primary Cor-lbl metric.

Abstract: Although probabilistic statements are ubiquitous, foundational disagreements persist about their understanding, as exemplified by debates between Bayesians and frequentists; moreover, it is unclear when and why acting on them actually leads to desirable outcomes. Here, we argue that every probability is the output of a \emph{prediction method}, that is, it depends on both a particular way of constructing abstractions and a way of transforming them into predictions. Through this, we provide a unifying perspective on supposedly different kinds of probabilities and show that even supposedly objective ones are model-dependent. We demonstrate that when a finite calibration criterion is met, one can anticipate the distribution of utilities for a given policy and inform successful decision-making on finite sets of events. Based on the notion of prediction methods, inductive arguments, and the probability calculus, we explain the feasibility of the calibration criterion in many settings. Overall, we develop a coherent perspective on probabilities and their use, connecting key intuitions behind other interpretations along the way.

Abstract: Long-term language-model agents rely on external memory across interactions. Atomic memories are particularly useful: their fine-grained semantic boundaries enable precise retrieval and direct comparison between observations. Yet accumulating atoms inevitably become redundant, overlapping, or conflicting. Existing methods often ask an LLM manager to add, update, delete, or rewrite memories directly, coupling semantic interpretation, storage decisions, and content generation in one error-prone operation. We introduce ROAM, a relation-guided framework that uses atomicity for management while allowing richer answer-time representations. ROAM classifies incoming--stored atom pairs as independent, equivalent, directionally subsuming, or conflicting, then organizes observations into active Primary and supporting Evidence roles. Fusion subsequently combines complementary details and temporal changes into compact, potentially non-atomic views. Only Primary views are retrieved for answering, preventing redundant or outdated atoms from competing independently. Across models and evaluation settings, ROAM improves answer accuracy by up to 29.8 percentage points. Ablations show complementary benefits from different relations and consistent gains from fusion beyond role organization. Mechanism analysis further finds 15.6-point higher answer-critical source recall and an 11.5-point lower confounder-token share. ROAM remains robust across manager scales.

Abstract: Frontier gains in language-model reasoning come from reinforcement learning on reasoning traces and are concentrated in domains with a cheap, sound verifier. We argue the field's binding constraint is the verification gap: no scalable, incorruptible reward for reasoning outside formal domains. We make four contributions. (1) Theory: in a joint-Gaussian model of best-of-N selection, verifier-gold correlation rho is the exact exchange rate between test-time compute and capability, and an unsound verifier pays a polynomial penalty N^(1/rho^2); a margin-free copula form predicts realized soundness of real LLM judges to 4% median error. (2) Demonstration: in program-synthesis testbeds with executable ground truth, including a pre-registered scaled replication, unsound verifiers lose Soundness-under-Pressure as optimization grows (0.94 to 0.32 at N=4096) while a sound verifier improves monotonically; reality-anchored settlement beats a frozen verifier under i.i.d. and adversarial pressure, driving the hacking gap from ~0.27 to ~0; soundness scales log-linearly with settled labels, with on-policy settlement ~10x more label-efficient than random labeling. With real LLM judges and unit-test execution as gold, a weak judge loses soundness under best-of-N (p<0.001), a stronger judge is more robust, and selection alone manufactures +0.53 hacking gaps from honest samples. Under real GRPO training, a frozen reward model traces the full overoptimization curve (executed reward collapses 90%) while the same model refit on a 10% settlement stream preserves 6x the executed reward. (3) Paradigm: proof-carrying cognition, where reasoning steps are typed probabilistic claims priced by a self-built world model trained only on held-out reality and settled by proper scoring rules. (4) Benchmark: we specify Soundness-under-Pressure as the headline metric for a reality-settled reasoning benchmark.

Abstract: Urban planning is a real-world spatial optimization problem that requires selecting feasible actions from large candidate spaces under practical objectives such as cost and service quality. Existing optimization and reinforcement learning methods are effective for fixed formulations, but often depend on task-specific representations and constraint handling. We propose \emph{CityPlanner}, a sandbox-agent framework for executable urban planning. CityPlanner introduces \emph{UrbanSandbox}, a unified file-based environment where agents inspect task files, generate plans, run evaluators, and revise decisions based on executable feedback. To make learning tractable, we further propose atomic-task reinforcement learning, which decomposes long sandbox trajectories into \emph{BuildPlan} for initial construction and \emph{ImprovePlan} for feedback-based refinement. Experiments on a real-world benchmark show that CityPlanner consistently outperforms heuristic, task-specific RL, and general LLM-agent baselines. Ablations verify the contributions of UrbanSandbox, atomic-task RL, and iterative deployment. We release the code and dataset at https://anonymous.4open.science/r/co-agent-C1C8

Abstract: Language models act through tools, yet practical agents face libraries containing thousands of interfaces. We introduce the tool menu as the short, ordered subset of available tools shown to an agent before execution. The agent can call only tools in this menu. Multi-step tasks require the final action and the prerequisite tools that create its inputs in a usable order. Current constructors rank tools by request relevance, which can surface the final action while omitting or delaying less obvious producers. We introduce the state path, a pre-execution route from the observable request state to the desired outcome, and propose State-Path Tool Menu to learn it. Our framework treats the menu as an execution prior over these routes. Its encoder represents which tools can run from the current state, how their outputs satisfy later inputs, and which orders recur in training paths. A retriever covers an executable entry, the missing-input producers, and the final action. A reranker then places producers before consumers. On ToolBench, our menu raises online success from 0.737 to 0.898 and outperforms retrieval, reranking, generation, and routing baselines without changing the agent. The State-Path menu also covers more complete chains with 32 tools than the official list covers with 128, and its success gain persists across executor families with different model capacities. Our code is at https://github.com/Met2348/State-Path.

Abstract: Large language models are increasingly deployed as agents that plan over long horizons and act through external tools. Most agents select actions through unconstrained generation over an accumulating history, leaving implicit the procedural knowledge of what to do, in what order, and under which conditions. As trajectories lengthen, agents can lose track of their objectives, invoke tools out of order, and repeat unproductive actions. We introduce the Procedural Graph: just as a knowledge graph organizes factual knowledge into (entity, relation, entity) triplets for what-is questions, a Procedural Graph organizes procedural knowledge into (procedure, relation, procedure) triplets for what-to-do questions. At each decision step, the framework localizes the agent's active node, and a guidance model translates the surrounding subgraph into step-level situational guidance that biases the solver's next action without dictating it. The graph is self-evolving: an LLM refiner contrasts failed trajectories with successful ones and edits the graph's topology and attributes, committing edits that preserve or improve held-out validation performance while retaining rejected ones to discourage repetition. Starting from a minimal skeleton, the loop builds graphs that match or surpass hand-designed ones. It can also repair a flawed expert prior. Across multiple datasets, task types, and LLMs, the Procedural Graph delivers consistent gains over memory-based baselines, and self-evolution further improves performance without manual engineering.

Abstract: High-quality tool-use data is critical for training language models to interact effectively with external tools. However, existing synthetic approaches typically follow a generate-then-filter paradigm with static post-hoc verification, often yielding inefficient data with imbalanced feature distributions. We propose ToolLoop, a closed-loop framework that decomposes synthesis into three progressive stages: (1) sampling function name combinations as ground truth; (2) backward derivation of user queries; and (3) forward derivation of tool calls. At each stage, dynamic self-feedback iteratively guides the model toward high-quality generation, realizing a transition from generate-then-filter to generate-verify-refine. On the Berkeley Function Calling Leaderboard (BFCL), a 4B parameter model trained with our 11K synthetic examples achieves 86.40% accuracy in non-reasoning mode, while an Isolate variant that removes BFCL-overlapping candidate functions still reaches 86.07\%. Cross-benchmark evaluation on ACEBench further demonstrates strong generalization, with 72.1% overall accuracy using only 18.3% of baseline training data.

Abstract: Let $P_t$ denote the induced path on $t$ vertices. Let $ω(G)$ denote the maximum number of vertices in a clique of a graph $G$. Previously Gyárfás (1987) proved that every $P_t$-free graph $G$ satisfies $χ(G)\le(t-1)^{ω(G)-1}$, and Gravier, Hoàng, and Maffray (2003) improved this to $χ(G)\le (t-2)^{ω(G)-1}$ for $t\ge4$. We prove that for $t\ge5$,every $P_t$-free graph $G$ satisfies $χ(G)<c_tλ_t^{ω(G)-1}$, where $λ_t=\tfrac12\bigl(t-2+\sqrt{t(t-4)}\bigr)<t-2$ and $c_t=\sqrt{1+4/(t(t-4))}=1+O(t^{-2})$ as $t\to\infty$. The proof is based on a refinement of the Gyárfás path argument and was found by Claude Fable 5.1 of Anthropic.

Abstract: We study graph-based automata: nondeterministic finite automata obtained from edge-colored or oriented graphs by taking every vertex as both initial and accepting, and every edge as a pair of opposite transitions. The language of these automata corresponds to the set of edge-colored or oriented paths mapping to their corresponding graphs. We develop an analogous notion for trees and characterise the languages recognised by these models. For tree languages we prove the existence of a unique size- minimal graph and, more generally, a homomorphism-minimal graph for both word and tree languages using duality methods. In order to further motivate these models, we showcase a few results at the inter- section between graph theory and automata theory: We relate graph-based automata to reversible automata, give a decomposition of graph-based languages into reversible languages, and introduce the remanent language of an undirected graph as the inter- section of all its orientations. This remanent language captures structural information on the graph such as chromatic number.

Abstract: Recent large language models can emit task-progress signals that agent frameworks use to decide whether a task should continue or stop, yet whether a model can reliably report its task progress at every stage of a task, and where and how its reports fail, has not been studied systematically. We evaluate this ability on the public benchmark $τ^2$-bench and on StageIF, a controlled testbed in which reporting checkpoints are placed across the task's lifecycle. Both settings require reports at multiple task stages. We find that reporting reliability depends on the stage a task has reached, and that almost every deployed model we test is reliable at some stages and unreliable at others. Where reporting breaks down is not the same everywhere. Most deployed models lose accuracy once work is under way and recover once the task is done. The newest generation closes that mid-task drop and instead grows conservative at the finish line. Our study exposes a capability gap in task-progress reporting and provides an evaluation protocol that spans the whole course of task execution for this ability on which agent operation depends. The findings indicate that agent frameworks should not control task flow on the strength of the model's state reports alone.

Abstract: Finite automata can describe relations of unbounded arity that are exponentially larger than their descriptions. We prove that constraint satisfaction for such relations is solvable in polynomial time whenever their length slices are preserved by a common fixed edge operation on a finite domain. The algorithm computes compact representations of the complete solution relation and its projections. Its main ingredient is a polynomial-time compilation of nondeterministic finite automata into the fork witnesses and small projections required by the few-subpowers algorithm. In the Mal'tsev case, a direct proof is polynomial also when the domain and operation table are supplied as input, answering the Mal'tsev tractability question for automatic constraint satisfaction. We also characterize all invariant relations of a family of 3-edge algebras with neither Mal'tsev nor near-unanimity terms. Their normal forms combine Boolean activity constraints with affine value spaces and yield canonical quadratic-bit representations constructible from NFAs or arbitrary generators. For graphoid automata, these results give polynomial-time recognition without a graph-width restriction, effective boundary composition, and comparison of finite graph relations. The quadratic boundary bounds are optimal in the worst case. A fixed three-state example separates polynomial-time recognition from hard exact counting.

Abstract: Multimodal continual learning has recently shown great potential for developing agents with human-like intelligence by continuously learning new tasks across multiple modalities. However, existing methods typically assume that the set of modalities per task is predefined and fixed. In this paper, we investigate a more realistic learning setting, referred to as dynamic multimodal continual learning, in which the set of modalities may vary across tasks rather than remaining fixed. This setting involves two primary challenges: (i) spatio-temporal catastrophic forgetting and (ii) adaptive multimodal fusion. To address these challenges, we propose NeuCME (as shorthand for \textbf{Neu}ral \textbf{C}ombinatorics of \textbf{M}ultiple \textbf{E}xperts), a novel framework designed to effectively learn and integrate knowledge across tasks with varying modalities. The proposed NeuCME model comprises three key components, namely modality-combinational rehearsal, multi-gated mixture-of-experts, and task relevance-guided distillation. Furthermore, we formulate an evaluation metric to quantify the dynamism of task sequences and then set up a comprehensive benchmark with different degrees of dynamism. Extensive experiments using four real-world datasets demonstrate that the proposed NeuCME outperforms state-of-the-art methods markedly.

Abstract: Large language models (LLMs) increasingly generate Markdown that is consumed by renderers, agents, code extractors, and structured downstream pipelines. Yet existing evaluations often conflate content quality with format adherence, leaving Markdown boundary failures under-measured. We introduce LatentMD, a benchmark and evaluation protocol for diagnosing CommonMark-level fence-boundary failures in LLM-generated Markdown. LatentMD separates content correctness from boundary correctness, enabling detection of outputs that are content-correct but boundary-broken. The benchmark contains 4,179 prompts and a CLI for scoring arbitrary model outputs. Across 9 LLMs and roughly 37,600 generations, we find that Markdown boundary failures are widespread: 38.0% of valid main-grid outputs are content-correct but boundary-broken, with substantial boundary breakage under unspecified prompts and in a small human-authored validation set. Ablations show that failures are driven primarily by same-family symmetric-delimiter collisions rather than nesting alone, are only partially mitigated by prompt hints, and generalize to Python triple-quote docstrings while JSON remains robust as an asymmetric-delimiter control. LatentMD provides a reproducible diagnostic target for parser-sensitive LLM evaluation.